Multi-agent system
A multi-agent system is a coordinated set of specialised AI agents that delegate to each other — each agent has a focused role, tool set, and system prompt rather than one mega-agent doing everything.
Multi-agent systems became a credible production pattern in 2026 because monolithic agents with 30+ tools mis-route badly. Frameworks: OpenAI Swarm, CrewAI, LangGraph hierarchies, Microsoft AutoGen, Claude sub-agents. Typical architectures: hierarchical (a coordinator + specialised workers), pipeline (chained specialists), or graph (state machine across many agents). Production reliability requires explicit handoff protocols, max-handoff caps, shared memory carefully scoped, and per-agent evals. Multi-agent is overkill for simple tasks — a single well-designed agent often beats a poorly-orchestrated multi-agent stack.
When to use multi-agent system
- Tasks with clear specialisation by sub-domain.
- When one agent's tool surface would exceed ~30.
Common mistakes
- Over-decomposing — too many agents add coordination cost.
- Skipping per-agent evals — the system fails in places that are hard to attribute.
FAQ
What is multi-agent system?
A multi-agent system is a coordinated set of specialised AI agents that delegate to each other — each agent has a focused role, tool set, and system prompt rather than one mega-agent doing everything.
When should I use multi-agent system?
Tasks with clear specialisation by sub-domain. When one agent's tool surface would exceed ~30.
What are the most common mistakes with multi-agent system?
Over-decomposing — too many agents add coordination cost. Skipping per-agent evals — the system fails in places that are hard to attribute.
Related terms
- AI agent — An AI agent is a system where a language model autonomously plans and executes a sequence of tool calls to accomplish a goal.
- Agent handoff — Agent handoff is the multi-agent pattern where one agent decides another specialised agent should take over the task — transferring the conversation state to the new agent's context.
- Agent loop — An agent loop is the repeating cycle of an AI agent — observe state, decide on an action (usually a tool call), execute, observe the result, and repeat — until a goal is reached or a stop condition fires.
- Mixture of agents — Mixture of agents is an inference pattern where multiple specialised LLM agents run in parallel and a router aggregator combines their outputs into a single answer — higher quality than any single agent at higher cost.
Last updated: 2026-06-01. Raw markdown: https://promtable.com/glossary/multi-agent.md.